The problem with AI is that, this time, nobody was caught off guard. From small startups to tech giants, everyone bet heavily on it. There are very good open-weight Chinese models, and even a French miracle. There is no winner-take-all.
So models are already a commodity, and will remain one for years. Companies must balance capex with very long payback periods against strict austerity everywhere else.
> From first principles I think youre right, but also Anthropic is making something like 80% margins.
I'm skeptical that they are operating at anything but a loss. 80% margins on a specific product (API) might still not cover the losses on the other products (Free tier, subscriptions) and their expenses (operating costs, training, research, etc).
Welp that’s not a good sign for the anthropic IPO. Curious what happens now that both OpenAI and Anthropic will face budgetary pressure. I don’t think they can raise prices, as it’d just open ppl to switch to Open source models.
It demonstrates why Anthropic has to IPO ASAP. Private investors are beginning to tap out. Anthropic is selling junior debt over 10% where it was about 8.5% last offer. If Anthropic wants to meet its compute purchase obligations (via actual business), it needs to start tapping public markets ASAP as its position is getting weaker. The size of the float and how much is primary vs secondary sale is the data we're missing right now. If the float is small and from secondary sales, the IPO is an exit and investors would be wise to stay away as Anthropic's runway stays where it is and therefore would require a re-acceleration of revenue growth to survive. If it is large and mostly from primary sales, it is a legit fund raise which extends the runway and revenue growth can be allowed to slow because of the pricing pressure you mentioned.
Shorting is more dangerous than buying long dated puts. At least with puts, if it's volatile there'll be opportunities to sell out of it if the direction isn't down in the long term.
It's so hard to time these things that short term bets are insanely risky. Yeah, thinks should go down, but they might or might not do it soon, and it might happen overnight because somethingsomething Yen Carry Trade or a surprise in the Middle East or some LatAm crash or an election result or God knows what.
The work going on to improve things on single card like a 3090 is nuts. HyperQwen, other things. I have my subscriptions, but always good to explore options.
You have to understand that Strata and HyperQwen are giving you good looking tokens per second but are not giving you anything close to the bf16 model’s capability (they quantize, worse they quantize KV caches, they approximate expert routing etc.)
The models are so impressive that getting 70-80% of their capability can be helpful, but beware, you are using a significantly dumber model.
I have subscriptions for necessary work and even pay for api calls.
The machine with the 3090 I built years ago and suddenly it’s worth 4x plus what I paid. It is for testing, but I use the model I run on it and the model I run on an mbp M1 Max with 64gig as workers for specifically scoped work.
I’d love to have a few dgx sparcs or the like to do heavy local stuff, but that money is best paying the pimps for now.
Sure, but, in absence of money, where are the open models going to come from? They take a lot of effort and resources to train. The current ones have been made in hope the growth will lead to a payoff one way or another. With the two main giants struggling and the market cooling, that payoff becomes quite uncertain.
could be stone soup. a bunch of companies pitch in to train an open-weight model, a consortium or something. I think we need a return to something like the GPL for it to work, though. otherwise companies will take a good model and post-train it without contributing.
Selling the same product. It doesn't matter who is running the neural network, the utility derived from the neural network has nothing to do with the name of the company.
No, they sell different products in the same category. You are basically saying Google and Bing as the same product when Bing is superior to Google search in some non-critical metrics.
Because they are? Google and Bing are the same product - search engines. Just like Ford and Ferrari both sell cars, whether you think one is "better" or not.
> No, they sell different products in the same category. You are basically saying Google and Bing as the same product when Bing is superior to Google search in some non-critical metrics.
In this case, it's pretty much the same product.
I mean, if you were to purchase electricity from two different providers (1x Nuclear plant and 1x Coal plant), would you actually care which one is being used to run your dishwasher on any given day?
I think Steve Jobs' quote sums this up beautifully:
"This is a field where one does not write a Principia which holds up for 200 years. This is not a field where one paints a painting that'll be looked at for centuries, or builds a church that will be admired and looked at in astonishment for centuries.
No, this is a field where one does one's work and in 10 years it's obsolete..."
And it's even harsher for AI where we're witnessing the release of a new model every hour these days
I think a lot of companies are struggling to move beyond this being a thing developers use to something everyone can use. Developers have to go really far out of their way to build agents and tools that apply the capabilities to their specific business. Someone in accounting cannot simply fire up chatgpt.com or copilot and use it in the same way a developer can. There is no instruction manual or historical precedent for how to do this. Very few people are currently in the right situation to make this work.
My observation is that most development teams are still in panic / survival mode, spending all day stacking sandbags instead of bringing these capabilities to the broader team. The fear of job loss in this particular market is crippling. No one will admit it publicly but you can feel the adversity on every AI status call. I've managed to convince a few that AI won't take their jobs, and these people have become indispensable at enabling deployment of AI beyond the weird nerd part of the org chart.
it’s kinda crazy but i do think we’re seeing a turn
my company was easily spending $1m/month on tokens
sonnet 5.5 drops - we barely cracked half this month
it has to go down too, carbon is interesting, china has qwen 4 soon
coding isn’t sustainable growth - RLRVF is really effective for software development - labs are cracking the code, these models will have to get better at other things to be worth it
How does this affect OpenAI's prospects in the next few quarters?
Unlike Anthropic, OAI does not have an S-1 to pore over. Their finances are shrouded in rumor and speculation. I can't assess how much of a death knell a ~15B financing shortfall is.
Nobody on the outside knows for sure, but two key signals to watch are project sunsets and brain drain.
If the rumors of fundraising difficulties are true, OAI will have to prioritize its most profitable projects, potentially slowing down new training runs and optimizing existing models for efficiency at the expense of user experience.
Insiders are also the first to sense internal trouble, so if top talent starts leaving for competitors preparing to go public, it won't bode well for the company.
The amount itself is probably meaningless. The fact that there was a shortfall at all is, for me, the signal.
Not too long ago, people were fighting to throw money at them. That might be the sign that we're past the peak of inflated expectations and going down towards the through of disillusionment .
> The amount itself is probably meaningless. The fact that there was a shortfall at all is, for me, the signal.
It is a signal, but I am not sure what it is signalling...
It could be signalling that all surplus money has been consumed already (i.e. the investment class has no more money left to invest).
It could signal that the specific class of investment is no longer desirable (i.e. the investment money is there, but has no appetite for token-providers).
It could mean that OpenAI, specifically, is considered not worth committing to anymore.
We should be looking at how the entire class of token providers is doing; if the rest are still getting interest from investors but OAI is not, then it's just OAI that is in trouble. If the entire class is not getting the money they are asking for, then maybe the peak of the bubble is reached (or the surplus money is all invested).
lack of appetite, when broadly visible, removes future speculators from the new investment pool, and starts bleeding old speculators. This will create a downward pressure. As the dynamic plays out the second rung of speculators will bleed out too.
Thats how bubbles burst - as each tranche of speculators bails the next tranche gets spooked through the pricing signals.
If this is being announced here, it was probably started at least a few months ago, and at that point funding was likely looking a lot more forthcoming, so it would make sense to try to raise more than is strictly necessary if the price is good. On that basis, it might not be a problematic shortfall.
If an open source model locally does one job correctly, lets say writing the code, then what really do we need OpenAI and Claude for? But I guess there might be some time in that.
That’s why the Mac mini and studio are selling like hotcakes. Massive unified memory with massive cpu performance, with a solid hardware. If I had 3-4K burning a hole in my pocket I know what I’d be getting.
What depreciation? I bought a DGX Spark a month ago for Qwen Flash Next (excellent model, by the way - using it basically exclusively at this point). The hardware has gone up in value 55% since I bought it. I just wish I had bought two of them.
I'll similarly ask, what depreciation? Hardware is going up in value, not down, and even if it did depreciate faster, people are optimizing models for older hardware so that again they retain more value for token generation.
Guess they'll just have to tighten the belt and follow through on their promise to revolutionize work, capitalism, government, and everything by building AGI. Lol. Lmao.
It's frightening to me that single companies' valuations can vary by that much. The stock exchange is the Adam Smith way of directing the assignation of resources to the most efficient company, and it's supposed to be liquid enough to allow fine-grained decisions.
> The directors with of such [joint-stock] companies, however, being the managers rather of other people's money than of their own, it cannot well be expected, that they should watch over it with the same anxious vigilance with which the partners in a private copartnery frequently watch over their own.... Negligence and profusion, therefore, must always prevail, more or less, in the management of the affairs of such a company.
This Adam Smith?
Adam Smith was against Joint Stock Companies (aka: the predecessor to the modern company). Specifically, he believed that shareholders will never watch over the money with the same care as a merchant who watches his own money.
> The stock exchange is the Adam Smith way of directing the assignation of resources to the most efficient company, and it's supposed to be liquid enough to allow fine-grained decisions.
So? OAI isn't on any stock exchange right now. Their value is not determined purely by supply and demand.
MSFT isn't laying off 60k people. It's offering voluntary severance, which will have a sub-10% take-up (for reference at GOOG it was low single digits).
I think Microsoft laying off 60,000 people was going to happen regardless of this AI boom or even the economy being in the absolute shitter it's in right now.
When the economy was much stronger in 2018 their headcount was around 131,000 people. By 2020 that jumped to 163,000, and by 2022 it reached 221,000. I have personally heard of employees feeling like they had nothing to do.
With their current head count at around 218,000 employees (I'm excluding their Activision head count here), bringing things back down to 160,000 or so employees sounds a bit more realistic, especially given that Microsoft has not been expanding its range of products significantly. They're building out more in the spaces they're in, but they've been pretty conservative on new product launches compared to others.
> Why do I sense a schadenfreude here? It's like people are rooting for their failure.
You sound surprised. Why?
After all, they have been selling their product on the basis of driving the value of human labour as close to zero as possible (your labour included). Why wouldn't people want such an endeavour to fail?
Among other reasons, because it's been evident that the numbers won't add up from the beginning. It's always delicious when reality catches on a major scam.
They're trying to enclose all of human knowledge so they can sell it back to us, all while displacing human labor and centralizing control of the economy.
They're telling us human endeavors are no longer needed if they succeeded, they also warn us that they might accidentally cause the end of the world in the process.
I’m not touching any of that stock, most likely ever.
There’s so much money that has been poured into this, with absolutely no revenue to justify it. Trillions of investments generating billions of losses.
Now they’re looking for retail and retirement funds to pay off the debt.
This is not to say that no one will make a profit; I’m sure some people are _very_ happy.
This is really fascinating, I'm thinking of putting a significant chunk of my net worth into the four AI labs (maybe 40/30/20/10% across Anthropic/OpenAI/Google/Meta, alongside another large investment across the chip supply chain).
It seems obvious to me that we will get superintelligence in the next 5 years, and that is a winner take all race. Who knows to whatever extent money will matter after this.
There is nothing to indicate that they will achieve general intelligence. Their models have stagnated and are incapable of continuous learning, so the architecture is a dead end.
The mistake is assuming that science is linear. You could give me a trillion dollars, there’s still no guarantee I’ll invent cold fusion. You can’t predict or guarantee inventions.
I don’t believe in betting on single companies but I’d be more inclined to bet on Amazon and Google. Most AI deployments go through bedrock, and Google has their own cloud business and other businesses to make money. That means they can win regardless of who develops the next cool model.
Interesting, the current gen of models feels quite close to AGI to me. If you showed it to me 5 years ago and told me nothing about how it works, I'd probably have called it AGI.
Maybe at first, but server hardware doesn't stay deployed tooo long and ends up liquidated. That was always a looming problem with this rapid buildout. Some hosts like OVH or Hetzner might swipe up some second hand gear, they're not afraid of renting old hardware. But eBay will be flooded once the DCs upgrade / downscale. And it wont be cheap at first either I bet. Everyone will try to cash in on the secondary market hype.
> No no no, don’t burst that AI bubble just yet. I haven’t finished all my side-projects.
You'll be fine; the GPUs will be sold off for pennies on the dollar, and if you can't run one at home there are plenty of people who will buy them to host for cost of hosting +5% markup.
Inference will keep getting cheaper and cheaper though. Perhaps model releases will be fewer and farther between because these groups can't affording paying out a billion or more in training each generation
Yeah, but Wright's law (cost of making the n^th something (like a token) being exponentially less than the cost of making the first) strongly suggests that post-bubble inference will be significantly cheaper the longer the bubble lasts.
Whether that's a good thing considering the other things those dollars could be going towards...
I mean I'm happy for better but it might start grounding the industry where 1.6% better on some benchmarks doesn't justify hundred million investments.
Over last 3 years on HN I have seen such low quality comments : prophesizing the end of times - the bubble will burst and we will all be so happy after that.
Any tiny news article that indicates even a modicum of challenges for OpenAI will be met with slurry of such comments.
Perhaps it’s good to keep a good epistemic hygiene in all this. OpenAI isn’t going away anywhere soon. AI isn’t going away, in fact we haven’t even started yet.
It's a reaction to the constant drum beat of doom marketing:
- "our product is too dangerous to release but we can't be bothered to prove its capabilities working in public hacking competitions"
- "our product is going to kill everyone"
- "our product is going to take all the jobs"
- "we have a $40T addressable market"
- "in a few days AI will be able to improve itself without human intervention and it's capabilities will hocky stick to infinity!!!"
Followed by absurd financials that spin massive losses as "getting in on the ground floor" before historic profits arrive based on the doom marketing above.
No one is saying AI is going away soon). But OpenAI and Anthro will go the way of Netscape when the market finally stops blindly accepting the doom marketing. Anyone with eyes can see LLMs will create historic increase in productivity among knowledge workers AND it will be a commodity. The massive profits frontier labs need to justify valuations are not coming.
There are AI bears and AI bulls. It's easy to cherry pick extreme and clearly wrong statements from each. But if you actually believe the bears are the more extreme/wrong overall you just aren't paying attention to what the bulls are saying.
At some level this all just comes off as post hoc justification for a really bad cortisol addiction.
Why are we keeping score in a game with completely arbitrary rules? The polarity is the hallmark sign for me of someone who enjoys a ride that certain neurotransmitter. I used to commute to and from work 2 hours every day. I know what it's like to be addicted to this stuff.
In other contexts people would ask me not to cherry pick extreme statements like “OpenAI and Anthropic will go away”. So I don’t have to cherry pick. I just have to pick what you said.
You have conflated two completely unrelated concepts. LLMs aren't going anywhere, but what good does that do a company burning hundreds of billions to do 20% better than Chinese models trained with millions? The economics don't economic.
It's far too late for a prescription now because they're already committed to hundreds of billions in spending they don't have, but the reason Chinese companies will win the long game is because they are serious and the Americans are deeply unserious. From the beginning, Chinese labs have viewed LLMs as software tools and developed efficiently. Conversely, Americans are selling a fantasy about creating the machine god if they just throw another hundred billion at it every few months. Altman, Amodei, and Musk are all pathological liars who have oriented their entire companies around deceiving investors and marketing hype rather than engineering principles. They are unbelievably successful at what they set out to do, but what they set out to do was not to build a long-lasting foundation for success.
They successfully turned the entire tech sector into a giant pyramid scheme. That's one hell of an accomplishment. I hope history forever remembers the names of everyone who aided and abetted them.
There's no should, only what economics incentivize. Right now we're not paying for the upfront training costs-it's great, the Millennial subsidy continues. But if, and big if right now, the infinite money font dries up and they have to recoup training investment direct from customers then that looks to be a huge price increase on the ground which makes cheap to train and good enough even more attractive than it already is.
Can also be read as observation, as I have observed that many colleagues haven't started yet using AI on a daily basis and I myself is still using methods from months ago (no agent orchestration).
Sure, but both can be true. The internet bubble bursting did not kill the internet. All that investment in severs and fibre was still there, and still got used. A lot of investors lost their shirts, but the Internet future still happened.
OpenAI going bust would not cause those data centres to cease to exist, they’d still be there just owned for cheap by someone else that can run them at cost.
A crash would drastically cut investment and research funding, but research would still happen. New companies would rise from the ashes of the old. It might still trigger a recession though.
> OpenAI going bust would not cause those data centres to cease to exist, they’d still be there just owned for cheap by someone else that can run them at cost.
Someone would have to see value in that. Right now the value prop of AI is that maybe someday it'll get better and we'll be able to build useful, profitable consumer products on it. That's not really a great investment thesis, unless you're drunk on the hype.
Now that the market is sobering up a bit, will operating a "cheap" GPU datacenter at cost be worth anything? Not "future value", but present value. Can you turn a healthy profit running a GPU datacenter today? I doubt it.
1. Guys so AI is a bubble but that doesn’t mean it’s not useful, just like dot com bubble
2. “Someone would have to see value in that. Right now the value prop of AI is that maybe someday it'll get better and we'll be able to build useful, profitable consumer products on it”
At this point AI company valuations only make sense if each of them captures most of the value of the entire IT sector. Soon. If that doesn’t happen, it’s a matter of time before reality catches up, but how much time? Who knows. It could be years still.
Or AI could in the next few years generate more value than the whole IT sector a few times over. That would be cool.
I think this is incorrect logic. OpenAI’s valuation doesn’t come from the sum of other tech company’s valuation.
You do know that all valuations can go up together right? It’s not zero sum. OpenAI can create trillions in value not by taking it from somewhere but by creating it from nothing.
LLMs not needed for programming or software engineering, and overall harmful in many dimensions. I look forward to the hype bubble popping and the AI industry collapsing.
Hopefully more of humanity's clearly vast amounts of discretionary wealth liquidity can be re-directed to better areas like health/medical, carbon crisis climate mitigation, clean/renewable energy, democracy defense and truly useful common infrastructure for society.
The problem with AI is that, this time, nobody was caught off guard. From small startups to tech giants, everyone bet heavily on it. There are very good open-weight Chinese models, and even a French miracle. There is no winner-take-all.
So models are already a commodity, and will remain one for years. Companies must balance capex with very long payback periods against strict austerity everywhere else.
Good news for customers, BTW.
> There are very good open-weight Chinese models
All the more reasons for us to figure out how to keep getting open models, or we are subject to the will of the the few frontier labs.
From first principles I think youre right, but also Anthropic is making something like 80% margins.
> From first principles I think youre right, but also Anthropic is making something like 80% margins.
I'm skeptical that they are operating at anything but a loss. 80% margins on a specific product (API) might still not cover the losses on the other products (Free tier, subscriptions) and their expenses (operating costs, training, research, etc).
Their S-1 says they're not operating at a loss. If they lied in the S-1, they'll go to jail for a decade or more.
You should probably read the S1, they showed a mostly nominal loss due to accounting rules.
They have a lot of cash, but very few assets and a voracious hunger for renting compute.
Don't get me wrong, they make a fantastic model, but the business success is yet to be seen.
the jail bit seems exceedingly unlikely
If you exclude all of their costs like hardware, training, $10m/engineer salaries, dojos, CEO sex islands…
GAAP are pretty clear that those things don't count, especially the last one.
What happens when the american deepseek equivalent comes online
Those are probably gross margins, not margins.
Welp that’s not a good sign for the anthropic IPO. Curious what happens now that both OpenAI and Anthropic will face budgetary pressure. I don’t think they can raise prices, as it’d just open ppl to switch to Open source models.
It demonstrates why Anthropic has to IPO ASAP. Private investors are beginning to tap out. Anthropic is selling junior debt over 10% where it was about 8.5% last offer. If Anthropic wants to meet its compute purchase obligations (via actual business), it needs to start tapping public markets ASAP as its position is getting weaker. The size of the float and how much is primary vs secondary sale is the data we're missing right now. If the float is small and from secondary sales, the IPO is an exit and investors would be wise to stay away as Anthropic's runway stays where it is and therefore would require a re-acceleration of revenue growth to survive. If it is large and mostly from primary sales, it is a legit fund raise which extends the runway and revenue growth can be allowed to slow because of the pricing pressure you mentioned.
I believe Anthropic IPO is a pump and dump at this point.
Play the market accordingly.
Short the stock at the first available opportunity --- if there is an opportunity.
Shorting is more dangerous than buying long dated puts. At least with puts, if it's volatile there'll be opportunities to sell out of it if the direction isn't down in the long term.
It's so hard to time these things that short term bets are insanely risky. Yeah, thinks should go down, but they might or might not do it soon, and it might happen overnight because somethingsomething Yen Carry Trade or a surprise in the Middle East or some LatAm crash or an election result or God knows what.
The work going on to improve things on single card like a 3090 is nuts. HyperQwen, other things. I have my subscriptions, but always good to explore options.
You have to understand that Strata and HyperQwen are giving you good looking tokens per second but are not giving you anything close to the bf16 model’s capability (they quantize, worse they quantize KV caches, they approximate expert routing etc.)
The models are so impressive that getting 70-80% of their capability can be helpful, but beware, you are using a significantly dumber model.
I have subscriptions for necessary work and even pay for api calls.
The machine with the 3090 I built years ago and suddenly it’s worth 4x plus what I paid. It is for testing, but I use the model I run on it and the model I run on an mbp M1 Max with 64gig as workers for specifically scoped work.
I’d love to have a few dgx sparcs or the like to do heavy local stuff, but that money is best paying the pimps for now.
Things like the hadamard rotations being applied before the quantisation mean that it has much less impact than before.
Thanks for the hardmard reference. It's pretty amazing what people can do with maths. I should have studied that instead of software.
Sure, but, in absence of money, where are the open models going to come from? They take a lot of effort and resources to train. The current ones have been made in hope the growth will lead to a payoff one way or another. With the two main giants struggling and the market cooling, that payoff becomes quite uncertain.
DOE, private labs, and universities are all working on building open models.
Here’s one example https://allenai.org/olmo
could be stone soup. a bunch of companies pitch in to train an open-weight model, a consortium or something. I think we need a return to something like the GPL for it to work, though. otherwise companies will take a good model and post-train it without contributing.
Could be state industry projects, maybe?
open weight, "how we trained it" is the open source of LLMs
real source is nice but boy have compile times gone up
Claude may have captured a different mindshare of the past year who tried AI before and it didn't work.
The more people see AI is still just software (albeit a different kind), the more they will remember to instead start with where/how it works.
[flagged]
Selling the same product. It doesn't matter who is running the neural network, the utility derived from the neural network has nothing to do with the name of the company.
No, they sell different products in the same category. You are basically saying Google and Bing as the same product when Bing is superior to Google search in some non-critical metrics.
Folks certainly have their preferences, but I think most consider each company's roughly equivalent models to be of the same quality.
It's more like a duopoly where the top two leaders have meaningful market share. Airbus and Boeing is more apt than Bing and Google.
I agree with this.
ChatGPT and Claude are differently operating and behaving products in the same space.
One is a little better/ different at considering the web, and the other is a juggernaut at sentences and reading in between / connecting the lines.
I'm also pretty proud, first flagged post ever for saying the obvious that they're two different companies.
They have different approaches to the tools they're providing. That fits people differently.
Because they are? Google and Bing are the same product - search engines. Just like Ford and Ferrari both sell cars, whether you think one is "better" or not.
[flagged]
Better comparison is Pepsi vs coke or Home Depot vs Lowe’s.
> No, they sell different products in the same category. You are basically saying Google and Bing as the same product when Bing is superior to Google search in some non-critical metrics.
In this case, it's pretty much the same product.
I mean, if you were to purchase electricity from two different providers (1x Nuclear plant and 1x Coal plant), would you actually care which one is being used to run your dishwasher on any given day?
I would also say that Google and Bing are roughly the same thing fwiw.
Like, no name of a company must or has to have anything to do with the product itself.
That's correct.
[flagged]
Please don't be snarky on HN. The guidelines make it clear we're trying for something better here. https://news.ycombinator.com/newsguidelines.html
I thought I had finished my sentence - the companies have different GTM, and maybe different trust/experience.
I think Steve Jobs' quote sums this up beautifully:
"This is a field where one does not write a Principia which holds up for 200 years. This is not a field where one paints a painting that'll be looked at for centuries, or builds a church that will be admired and looked at in astonishment for centuries.
No, this is a field where one does one's work and in 10 years it's obsolete..."
And it's even harsher for AI where we're witnessing the release of a new model every hour these days
I think a lot of companies are struggling to move beyond this being a thing developers use to something everyone can use. Developers have to go really far out of their way to build agents and tools that apply the capabilities to their specific business. Someone in accounting cannot simply fire up chatgpt.com or copilot and use it in the same way a developer can. There is no instruction manual or historical precedent for how to do this. Very few people are currently in the right situation to make this work.
My observation is that most development teams are still in panic / survival mode, spending all day stacking sandbags instead of bringing these capabilities to the broader team. The fear of job loss in this particular market is crippling. No one will admit it publicly but you can feel the adversity on every AI status call. I've managed to convince a few that AI won't take their jobs, and these people have become indispensable at enabling deployment of AI beyond the weird nerd part of the org chart.
it’s kinda crazy but i do think we’re seeing a turn
my company was easily spending $1m/month on tokens
sonnet 5.5 drops - we barely cracked half this month
it has to go down too, carbon is interesting, china has qwen 4 soon
coding isn’t sustainable growth - RLRVF is really effective for software development - labs are cracking the code, these models will have to get better at other things to be worth it
We’ve seen a similar drop in spend. 3-5x less for entire teams comparing September and October.
It’s not clear whether people are being more careful, or if the price per effort is in free fall.
Since Sol 6.1 came out, I have come nowhere close to my daily token budget. Previously, it was basically every day.
Would it partly be explained by Haiku 5.5?
At work I am stuck with Claude, but I was always mindful of the cost of that thing.
With aggressive clean session management and some judicious compacting, I get Opus to plan, Haiku to code, Sonnet to review.
I managed to reduce costs from ~$200 to ~$50 in a month, even when I started to use Claude for more tasks than I did before.
1m/month? whoever is running your company is out to lunch.
~2000 engineers with low adoption
1970 engineers too many
[dead]
How does this affect OpenAI's prospects in the next few quarters?
Unlike Anthropic, OAI does not have an S-1 to pore over. Their finances are shrouded in rumor and speculation. I can't assess how much of a death knell a ~15B financing shortfall is.
Nobody on the outside knows for sure, but two key signals to watch are project sunsets and brain drain.
If the rumors of fundraising difficulties are true, OAI will have to prioritize its most profitable projects, potentially slowing down new training runs and optimizing existing models for efficiency at the expense of user experience.
Insiders are also the first to sense internal trouble, so if top talent starts leaving for competitors preparing to go public, it won't bode well for the company.
The amount itself is probably meaningless. The fact that there was a shortfall at all is, for me, the signal.
Not too long ago, people were fighting to throw money at them. That might be the sign that we're past the peak of inflated expectations and going down towards the through of disillusionment .
> The amount itself is probably meaningless. The fact that there was a shortfall at all is, for me, the signal.
It is a signal, but I am not sure what it is signalling...
It could be signalling that all surplus money has been consumed already (i.e. the investment class has no more money left to invest).
It could signal that the specific class of investment is no longer desirable (i.e. the investment money is there, but has no appetite for token-providers).
It could mean that OpenAI, specifically, is considered not worth committing to anymore.
We should be looking at how the entire class of token providers is doing; if the rest are still getting interest from investors but OAI is not, then it's just OAI that is in trouble. If the entire class is not getting the money they are asking for, then maybe the peak of the bubble is reached (or the surplus money is all invested).
lack of appetite, when broadly visible, removes future speculators from the new investment pool, and starts bleeding old speculators. This will create a downward pressure. As the dynamic plays out the second rung of speculators will bleed out too.
Thats how bubbles burst - as each tranche of speculators bails the next tranche gets spooked through the pricing signals.
If this is being announced here, it was probably started at least a few months ago, and at that point funding was likely looking a lot more forthcoming, so it would make sense to try to raise more than is strictly necessary if the price is good. On that basis, it might not be a problematic shortfall.
Nah, they were planning for 30bn two weeks ago, so this is a bad look if nothing else.
They’re dead. As soon as people start getting fearful you’ll go to zero, they won’t have FOMO anymore. FOMO was the only reason people were investing.
oh but that's nothing... consider all previous financing episodes and the companies involved
If an open source model locally does one job correctly, lets say writing the code, then what really do we need OpenAI and Claude for? But I guess there might be some time in that.
That’s why the Mac mini and studio are selling like hotcakes. Massive unified memory with massive cpu performance, with a solid hardware. If I had 3-4K burning a hole in my pocket I know what I’d be getting.
You won't. It will become something you install like a browser.
The depreciation on a sufficiently powerful local machine costs more than the subscription to cloud frontier AI.
What depreciation? I bought a DGX Spark a month ago for Qwen Flash Next (excellent model, by the way - using it basically exclusively at this point). The hardware has gone up in value 55% since I bought it. I just wish I had bought two of them.
I'll similarly ask, what depreciation? Hardware is going up in value, not down, and even if it did depreciate faster, people are optimizing models for older hardware so that again they retain more value for token generation.
There is no depreciation on memory these days, heck ddr4 and ddr3 even are resurging to 10x pricing. Ram is an investment vehicle this year.
Also spinning hdds… insane amounts of appreciation.
Wow! Its bursting...
The subheading in the article is telling, and doesn't appear to demonstrate anything I would describe as "bursting"
> The ChatGPT maker’s latest funding round values it at $1.25 trillion, less than its $1.4 trillion target
The investments these companies are making depends entirely on the confidence that number go up, a lot, very quickly and soon.
Number go down, or even not go up as much as expected, is not good. These valuations are entirely matters of confidence. When that goes away…
Guess they'll just have to tighten the belt and follow through on their promise to revolutionize work, capitalism, government, and everything by building AGI. Lol. Lmao.
That's 150B shaved off like it's nothing lol
Is there even a way to professionally decide if a company's value is 1.25 or 1.4 Trillion?
I always kind of suspected that those are somehow derived numbers, but arbitrary for the most part.
Its based on the price of the most recent round, if I just sold 10% for 1 million, then naively 100% is worth 10 million.
The implication is they had to sell more of the company for cheaper than they wanted if they only hit 1.25 instead of 1.4.
It's frightening to me that single companies' valuations can vary by that much. The stock exchange is the Adam Smith way of directing the assignation of resources to the most efficient company, and it's supposed to be liquid enough to allow fine-grained decisions.
> The directors with of such [joint-stock] companies, however, being the managers rather of other people's money than of their own, it cannot well be expected, that they should watch over it with the same anxious vigilance with which the partners in a private copartnery frequently watch over their own.... Negligence and profusion, therefore, must always prevail, more or less, in the management of the affairs of such a company.
This Adam Smith?
Adam Smith was against Joint Stock Companies (aka: the predecessor to the modern company). Specifically, he believed that shareholders will never watch over the money with the same care as a merchant who watches his own money.
They’re not a public company so the only people buying in are insiders and the rich.
> The stock exchange is the Adam Smith way of directing the assignation of resources to the most efficient company, and it's supposed to be liquid enough to allow fine-grained decisions.
So? OAI isn't on any stock exchange right now. Their value is not determined purely by supply and demand.
[flagged]
Not looking very good fellas, Bot Swarm breaches, microslop laying off 27% of their workforce.
If you have a billion users and 90% + are freeloaders, you need social media style ads to support growth, which no one likes.
Plus they are all out of data training sets, they are buying zombie companies for their data.
There are dozens of distillated clones, which are getting better every release.
The only moat is expensive data centre builds with on-loan Nvidia inference and the Chinese embargo fears.
Good luck out there.
MSFT isn't laying off 60k people. It's offering voluntary severance, which will have a sub-10% take-up (for reference at GOOG it was low single digits).
I think Microsoft laying off 60,000 people was going to happen regardless of this AI boom or even the economy being in the absolute shitter it's in right now.
When the economy was much stronger in 2018 their headcount was around 131,000 people. By 2020 that jumped to 163,000, and by 2022 it reached 221,000. I have personally heard of employees feeling like they had nothing to do.
With their current head count at around 218,000 employees (I'm excluding their Activision head count here), bringing things back down to 160,000 or so employees sounds a bit more realistic, especially given that Microsoft has not been expanding its range of products significantly. They're building out more in the spaces they're in, but they've been pretty conservative on new product launches compared to others.
The market has chosen to crown $ANT as queen.
What else can you expect? Look the markets right now. Its bleeding. Started with China and now US.
No archive.{today,is,ph,md}, no wallhop.io, no Javascript, no User-Agent heaader, no cookies, HTTPS optional
Text-only (simple HTML)
Text-only (JSON)https://assets.msn.com/content/view/v2/Detail/en-in/AA2grM9h
Why do I sense a schadenfreude here? It's like people are rooting for their failure.
> Why do I sense a schadenfreude here? It's like people are rooting for their failure.
You sound surprised. Why?
After all, they have been selling their product on the basis of driving the value of human labour as close to zero as possible (your labour included). Why wouldn't people want such an endeavour to fail?
> driving the value of human labour as close to zero as possible
Tech and automation exists to drive the marginal cost of labor toward zero. It's just happening faster now.
> (your labour included)
Right, it's disrupting you now instead of just dumb people.
This is false. Other forms of tech were not designed to replace humans with some kind of successor species, as they're trying to do here.
Other tech was developed as a productivity boost, and for jobs it may have displaced, it ultimately created more.
I expect people here to root for the future. In the future AI overtaking humans is inevitable.
The ordeal feels like 2yo who hasn't learned yet that the pee will come no matter if he wishes to play or not.
They've been rooting for ours, it's only fair.
Among other reasons, because it's been evident that the numbers won't add up from the beginning. It's always delicious when reality catches on a major scam.
They're trying to enclose all of human knowledge so they can sell it back to us, all while displacing human labor and centralizing control of the economy.
They're telling us human endeavors are no longer needed if they succeeded, they also warn us that they might accidentally cause the end of the world in the process.
Why shouldn't we root for their failure?
Some of us truly are.
this is more about the supply and demand of capital in any given moment than it is about the outlook of the company.
as a litmus test: would you turn down the opportunity to buy OpenAI shares at this current valuation?
I’m not touching any of that stock, most likely ever.
There’s so much money that has been poured into this, with absolutely no revenue to justify it. Trillions of investments generating billions of losses.
Now they’re looking for retail and retirement funds to pay off the debt.
This is not to say that no one will make a profit; I’m sure some people are _very_ happy.
This is really fascinating, I'm thinking of putting a significant chunk of my net worth into the four AI labs (maybe 40/30/20/10% across Anthropic/OpenAI/Google/Meta, alongside another large investment across the chip supply chain).
It seems obvious to me that we will get superintelligence in the next 5 years, and that is a winner take all race. Who knows to whatever extent money will matter after this.
There is nothing to indicate that they will achieve general intelligence. Their models have stagnated and are incapable of continuous learning, so the architecture is a dead end.
The mistake is assuming that science is linear. You could give me a trillion dollars, there’s still no guarantee I’ll invent cold fusion. You can’t predict or guarantee inventions.
I don’t believe in betting on single companies but I’d be more inclined to bet on Amazon and Google. Most AI deployments go through bedrock, and Google has their own cloud business and other businesses to make money. That means they can win regardless of who develops the next cool model.
Interesting, the current gen of models feels quite close to AGI to me. If you showed it to me 5 years ago and told me nothing about how it works, I'd probably have called it AGI.
How/why is SI winner-take-all?
Total power over others. If they want to maintain it, they can just make it so and nobody can stop them.
yes
> you turn down the opportunity to buy OpenAI shares at this current valuation?
1.25T? Yes.
Open models exist, and are extremely capable. OpenAI will never achieve a decent ROI when that is true.
MSFT alone is 4T and is probably at least 75% income-from-software.
If either of AI companies "solves software" (does not look improbable rn) ahead of all others, that would defend ~3T valuation right away.
And that's just software.
1) I don't believe that will happen.
2) Even if does, Chinese open models are very close in capabilities.
I thought Masayoshi Son would step in to back them?
The end state is OpenAI as a division of Microsoft.
At this rate ANT will end up as a division of Amazon
I always pegged Apple for ANT.. But I bet Apple is going to poach talent from Anthropic post IPO after they got their payout and are ready to move on.
> I always pegged Apple for ANT..
Those two are very different cults. Cultures, I mean.
I like what you did there
There is absolutely no timeline where Apple acquires any token provider because Apple doesn't sell tokens.
Anthropic's assets aren't tokens.
Is this supposed to be clever? Their assets serve their business (you know ... like any normal company) and that business is selling tokens.
The new Siri is Gemini rebranded, so Apple does sell tokens, they just charge for it as a bundle with the hardware.
> The new Siri is Gemini rebranded
No. It’s actual models from Apple, trained from gemini outputs, which is not the same as reselling gemini.
hn is purely a rumor mill now (you are correct).
I mean the super groups are going to break up and make their own bands.
My only request is to keep Opus 5.5 alive
the liquidity is drying out.
No no no, don’t burst that AI bubble just yet. I haven’t finished all my side-projects.
Once the bubble bursts we can finally sweep up the datacenter GPUs and jerry rig them to our home servers!
I literally can't wait.
[dead]
The top models still require 1T+ of VRAM. Even if prices go down, that's still gonna be very unaffordable.
They will probably go to other data centers or company premises.
Maybe at first, but server hardware doesn't stay deployed tooo long and ends up liquidated. That was always a looming problem with this rapid buildout. Some hosts like OVH or Hetzner might swipe up some second hand gear, they're not afraid of renting old hardware. But eBay will be flooded once the DCs upgrade / downscale. And it wont be cheap at first either I bet. Everyone will try to cash in on the secondary market hype.
After you upgrade your electric panel and cooling.
Winter is coming.
The government will just nationalize them and use all the compute for poorly defined military tasks
Or to further build out total domestic surveillance, like they've been telegraphing.
Use Ultra mode.
> No no no, don’t burst that AI bubble just yet. I haven’t finished all my side-projects.
You'll be fine; the GPUs will be sold off for pennies on the dollar, and if you can't run one at home there are plenty of people who will buy them to host for cost of hosting +5% markup.
The consumer-level services you've gotten used to will still exist no matter how hard the bubble bursts, if it ever does. Way too many players now.
Sure, but you'll have to pay for what it really costs
If the capital cost gets written off, because it can never be recouped, then there’s only the operating cost left.
Suppose OpenAI goes bust tomorrow and all those data centres get sold for pennies on the dollar. Whoever buys them can operate them at cost.
I thought the incremental supply cost per token wasn’t that high, and the main expense for these companies was the training they do on top of that?
Did internet get more expensive after the dot com bursted?
Inference will keep getting cheaper and cheaper though. Perhaps model releases will be fewer and farther between because these groups can't affording paying out a billion or more in training each generation
Yeah, but Wright's law (cost of making the n^th something (like a token) being exponentially less than the cost of making the first) strongly suggests that post-bubble inference will be significantly cheaper the longer the bubble lasts.
Whether that's a good thing considering the other things those dollars could be going towards...
The chinese models are being trained for less than a million a pop.
I mean I'm happy for better but it might start grounding the industry where 1.6% better on some benchmarks doesn't justify hundred million investments.
I have only ever paid per token for this reason: if I can’t generate sufficient value to cover the real cost I don’t want to invest in using it
.. which will be the same you are paying today. Token prices are already above their operating costs.
If you're talking about free tiers, yes, that might change as marketing budgets will shrivel up.
Over last 3 years on HN I have seen such low quality comments : prophesizing the end of times - the bubble will burst and we will all be so happy after that.
Any tiny news article that indicates even a modicum of challenges for OpenAI will be met with slurry of such comments.
Perhaps it’s good to keep a good epistemic hygiene in all this. OpenAI isn’t going away anywhere soon. AI isn’t going away, in fact we haven’t even started yet.
It's a reaction to the constant drum beat of doom marketing:
- "our product is too dangerous to release but we can't be bothered to prove its capabilities working in public hacking competitions" - "our product is going to kill everyone" - "our product is going to take all the jobs" - "we have a $40T addressable market" - "in a few days AI will be able to improve itself without human intervention and it's capabilities will hocky stick to infinity!!!"
Followed by absurd financials that spin massive losses as "getting in on the ground floor" before historic profits arrive based on the doom marketing above.
No one is saying AI is going away soon). But OpenAI and Anthro will go the way of Netscape when the market finally stops blindly accepting the doom marketing. Anyone with eyes can see LLMs will create historic increase in productivity among knowledge workers AND it will be a commodity. The massive profits frontier labs need to justify valuations are not coming.
There are AI bears and AI bulls. It's easy to cherry pick extreme and clearly wrong statements from each. But if you actually believe the bears are the more extreme/wrong overall you just aren't paying attention to what the bulls are saying.
At some level this all just comes off as post hoc justification for a really bad cortisol addiction.
Why are we keeping score in a game with completely arbitrary rules? The polarity is the hallmark sign for me of someone who enjoys a ride that certain neurotransmitter. I used to commute to and from work 2 hours every day. I know what it's like to be addicted to this stuff.
In other contexts people would ask me not to cherry pick extreme statements like “OpenAI and Anthropic will go away”. So I don’t have to cherry pick. I just have to pick what you said.
> OpenAI isn’t going away anywhere soon. AI isn’t going away
You have conflated two completely unrelated concepts. LLMs aren't going anywhere, but what good does that do a company burning hundreds of billions to do 20% better than Chinese models trained with millions? The economics don't economic.
> Chinese models trained with millions
I see you are still under illusion that Chinese models are trained with millions.
According to your argument, OpenAI should simply stop R and D which can result in massive jumps in performance because China may catch up?
What is your prescription if I may ask? Don’t invest in research? Let China do it?
It's far too late for a prescription now because they're already committed to hundreds of billions in spending they don't have, but the reason Chinese companies will win the long game is because they are serious and the Americans are deeply unserious. From the beginning, Chinese labs have viewed LLMs as software tools and developed efficiently. Conversely, Americans are selling a fantasy about creating the machine god if they just throw another hundred billion at it every few months. Altman, Amodei, and Musk are all pathological liars who have oriented their entire companies around deceiving investors and marketing hype rather than engineering principles. They are unbelievably successful at what they set out to do, but what they set out to do was not to build a long-lasting foundation for success.
They successfully turned the entire tech sector into a giant pyramid scheme. That's one hell of an accomplishment. I hope history forever remembers the names of everyone who aided and abetted them.
Calling OpenAI and Anthropic a pyramid scheme is absurd
There's no should, only what economics incentivize. Right now we're not paying for the upfront training costs-it's great, the Millennial subsidy continues. But if, and big if right now, the infinite money font dries up and they have to recoup training investment direct from customers then that looks to be a huge price increase on the ground which makes cheap to train and good enough even more attractive than it already is.
[flagged]
Sometines it truly feels like titles of YouTube videos.
You know "we haven't even started yet" is also prophesying, right?
Can also be read as observation, as I have observed that many colleagues haven't started yet using AI on a daily basis and I myself is still using methods from months ago (no agent orchestration).
Sure, but both can be true. The internet bubble bursting did not kill the internet. All that investment in severs and fibre was still there, and still got used. A lot of investors lost their shirts, but the Internet future still happened.
OpenAI going bust would not cause those data centres to cease to exist, they’d still be there just owned for cheap by someone else that can run them at cost.
A crash would drastically cut investment and research funding, but research would still happen. New companies would rise from the ashes of the old. It might still trigger a recession though.
> OpenAI going bust would not cause those data centres to cease to exist, they’d still be there just owned for cheap by someone else that can run them at cost.
Someone would have to see value in that. Right now the value prop of AI is that maybe someday it'll get better and we'll be able to build useful, profitable consumer products on it. That's not really a great investment thesis, unless you're drunk on the hype.
Now that the market is sobering up a bit, will operating a "cheap" GPU datacenter at cost be worth anything? Not "future value", but present value. Can you turn a healthy profit running a GPU datacenter today? I doubt it.
> That's not really a great investment thesis, unless you're drunk on the hype.
To be fair it is very visibly getting better basically nonstop in the last 2 years at least.
At the risk of goomba fallacy
1. Guys so AI is a bubble but that doesn’t mean it’s not useful, just like dot com bubble
2. “Someone would have to see value in that. Right now the value prop of AI is that maybe someday it'll get better and we'll be able to build useful, profitable consumer products on it”
[dead]
It is the α's and Z's brigading in with their over-emotional shortsighted groupthink.
1. GPT-4.5 launch. "AI crash is gonna lead to decade long winter." https://news.ycombinator.com/item?id=43206226
2. Claude Code limits tightened. A commenter expects the bubble to pop, while arguing AI will succeed long term. https://news.ycombinator.com/item?id=44599297
3. GPT-5 disappoints. "Teetering on the edge of a huge bubble pop." https://news.ycombinator.com/item?id=44966583
4. DeepSeek releases a cheaper model. A commenter expects excess data-centre capacity when the bubble bursts. https://news.ycombinator.com/item?id=45413548
5. AMD signs an OpenAI chip deal. A commenter predicts a bubble pop and a flood of cheap GPUs. https://news.ycombinator.com/item?id=45491514
6. Discussion of AI-assisted open-source contributions. A commenter says models must improve soon or the bubble will pop. https://news.ycombinator.com/item?id=45745326
7. Discussion of Opus 4.5 coding. A commenter advises using cheap AI tools before the bubble bursts. https://news.ycombinator.com/item?id=46521192
8. GitHub Copilot moves to usage-based billing. "The bubble has burst." https://news.ycombinator.com/item?id=47929404
9. DeepSeek runs on one AMD accelerator. A commenter predicts a bubble pop will flood the hardware market. https://news.ycombinator.com/item?id=49166790
10. Same DeepSeek thread. "AI is overvalued, so it WILL pop." https://news.ycombinator.com/item?id=49167787
Edit: more examples
1. AI on an old CPU: coding gains don’t justify investment; bubble will burst. https://news.ycombinator.com/item?id=48938412
2. Nvidia–OpenAI deal stalls: GPU bubble, inflated demand. https://news.ycombinator.com/item?id=46834749
3. Lisp machines retrospective: AI bankruptcies and another bubble. https://news.ycombinator.com/item?id=45990309
4. Nvidia customer concentration: AI bubble will pop despite useful technology. https://news.ycombinator.com/item?id=41411258
5. DeepSeek open-source announcement: inflated AI valuations will collapse. https://news.ycombinator.com/item?id=43126535 https://news.ycombinator.com/item?id=43126619
6. OpenAI–AWS deal: spending commitments could bring everything crashing down. https://news.ycombinator.com/item?id=45802550
7. Anthropic funding: final big push before the bubble bursts. https://news.ycombinator.com/item?id=45105333
8. Enterprise AI adoption: disappointing returns will pop the bubble within two years. https://news.ycombinator.com/item?id=48025025
9. OpenAI cash burn: railroad-style bubble and massive investment downturn. https://news.ycombinator.com/item?id=46439964
OOC did you have this list prepared? You posted it very quickly as a response
Maybe it’s their day job… they seem to be the kind of commenter that doesn’t have any interest outside of LLMs.
Like I said, brigading botlike behavior.
At this point AI company valuations only make sense if each of them captures most of the value of the entire IT sector. Soon. If that doesn’t happen, it’s a matter of time before reality catches up, but how much time? Who knows. It could be years still.
Or AI could in the next few years generate more value than the whole IT sector a few times over. That would be cool.
I think this is incorrect logic. OpenAI’s valuation doesn’t come from the sum of other tech company’s valuation.
You do know that all valuations can go up together right? It’s not zero sum. OpenAI can create trillions in value not by taking it from somewhere but by creating it from nothing.
You kind of proved his point about brigading with this giant link dump.
ruh roh
LLMs not needed for programming or software engineering, and overall harmful in many dimensions. I look forward to the hype bubble popping and the AI industry collapsing.
Hopefully more of humanity's clearly vast amounts of discretionary wealth liquidity can be re-directed to better areas like health/medical, carbon crisis climate mitigation, clean/renewable energy, democracy defense and truly useful common infrastructure for society.
Is anybody missing the Pelicans posts at the top of the AI threads? I don't.
I haven't read that article, but looking at WSJ, I'd take it with a pinch of salt. It's not a reliable source.
[dead]